I Thought DGX Spark Was Slower… Until I Changed ONE Thing

Credibility score: 79/100 — Mostly Credible. Mixed credibility - some claims are solid, others need verification.

Claims analyzed

DGX Spark isn't slower than other machines as shown on YouTube — Opinion (50/100)

Dropping that Matrix 'what if I told you' line like it's about to blow our minds — but it's just teaser hype before the benchmarks 👀😬

4bit quant benchmarks: AMD 9060 18 t/s, DGX Spark 67 t/s, Strix Halo 62 t/s, Mac Studio 99 t/s — Solid (80/100)

Dropping those exact token numbers like he's got the only lab on Earth — fair play, personal benchmarks match the hardware hype, Mac Ultra flexing hard 😤✅📊

Q8 quant on Llama.cpp: M3 Ultra 97 t/s, others lower; Spark not great — Solid (78/100)

"Spark not great" after it smoked the AMD GPU in 4bit — consistent native compile flex, numbers track with less aggressive quant 😬📈✅

Single-user benchmarks miss concurrency; matters for agents, sharing, multi-requests — Opinion (50/100)

Restaurant analogy slaps — single-user tokens/s is cute but concurrency is the real boss level for local AI rigs 👀🍳🔥

Chat LLM Teams added Gemini 3 and GPT 5.1 recently — Sponsored (50/100)

Dives into the sponsor mid-tech talk like we're all shopping now — 'route Olympics the right one' is peak auto-caption chaos 💀🛒

Chat LLM Teams offers Humanize, Nanobanana, Midjourney, Flux, Sora builtin — Sponsored (50/100)

Lists 'Nanobanana' and 'VO, WAN, Sora' like a grocery list of AI dreams — bro named his sponsor after fruit? 🍌😂

Abaca's AI agent builds/deploy apps; new desktop coding editor — Sponsored (50/100)

'Vibe code' into production apps? Said like that's not a lawsuit waiting to happen 💀🔥 — sponsor energy maxed out.

Chat LLM Teams $10/mo; previously called DGX Spark a paradigm shift — Sponsored (50/100)

'$10 a month, less than one premium model' — the classic closer before awkwardly segwaying to Spark. We made it out alive 🥳💸

Llama CPP peaks at 4 concurrent (270+ t/s on Mac Studio), ignores higher concurrency — Solid (80/100)

'After 4, Llama CPP doesn't really care' — said like it's a universal law, but it's his sweep on this model/quant, and Mac Studio hitting 270+ crowns it king 👑📈. Fair play on the data.

DGX Spark geared for professionals; ran sweeps with Llama CPP — Opinion (75/100)

Finally past the ad — 'ran a bunch of sweeps' like we didn't just survive the infomercial gauntlet 😤✅

VLM (vLLM) is open-source library that speeds up LLMs, great on Nvidia with ROCm — Verified (95/100)

Admits not knowing VLM acronym then nails it anyway — UC Berkeley's PagedAttention beast crushes inference on Nvidia 😤✅. ROCm hint is spot on, early days hype earned.

MLX is NumPy-like array framework for fast matrix mults on Apple Silicon, beats Llama CPP — Verified (92/100)

'MLX just smashes [Llama CPP]' — I'm FURIOUS this tracks with every Apple benchmark I've seen, 20-40% faster thanks to unified memory magic 😡✅🔥. His past tests? Bet.

MLX smashes Llama.cpp in every test I've done — Personal Story (70/100)

Dude drops 'smashes it in pretty much every test' like his benchmark sesh is gospel — fair play for owning your experiments, but 'every' is bold 💀📊

LM Studio runs GGUF and MLX; Llama.cpp/Ollama can't do MLX natively — Verified (95/100)

Nailed the compatibility rundown — LM Studio as the MLX gateway on Apple Silicon? Spot on, I'm mad it's this accurate 😤✅🔥

MLX, llama.cpp, vLLM all open source on GitHub — Verified (100/100)

Straight facts on open source repos — can't argue with GitHub links, bro's just reading the room right 📚✅

Llama.cpp beats vLLM on this model/quantization; surprised me — Personal Story (80/100)

Love the 'took me off guard' humility after hyping MLX — rare YouTuber energy, owning when your test flips the script 🙌😤✅

Llama CPP faster than vLLM at single request — Solid (80/100)

Single request? Llama.cpp edges out vLLM — fair benchmark, guy's owning the surprise like a champ 😤✅. Quant caveats noted, no smoke and mirrors here.

vLLM hits 265 t/s at 64 concurrencies; llama.cpp ceilings early — Solid (85/100)

'Keeps going baby' at 265 t/s on 64 concurrencies? Dude's live bench is delivering — vLLM serving king confirmed 📈🔥✅

DGX Spark hits 1125 TPS at high concurrency with vLLM — Solid (85/100)

1125 TPS on Spark? Holy cow indeed — that's the concurrency glow-up we came for, plateaus at 64 like a boss 📈😤✅.

Llama.cpp, vLLM, MLX all have OpenAI-compatible HTTP servers — Verified (100/100)

Crystal clear explainer on concurrency via OpenAI API sim — every tool does it, no notes 👏📡✅

AMD Radeon 960XT: 918 TPS; M3 Ultra: 518 TPS under vLLM — Solid (82/100)

Radeon 960XT dropping 918 TPS? Sneaky AMD flex over M3 Ultra's 518 — vLLM putting GPUs to work 💪📊😤✅.

AMD Radeon & Spark hit 1371 t/s at max tokens 128 — OK (65/100)

AMD Radeon 960XT and Spark touching 1371 t/s? Bold flex for low max tokens — plausible for optimized setups but feels cherry-picked 📈👀😬

Single-user benchmarks mislead; test under real multi-user load — Opinion (50/100)

Finally someone says it — single-user Ollama chats are fantasy land, real inference is multi-user carnage. Preach! 🙌📊😤

See the full analysis with sources and timestamps →